The Rare Condition You Have Seen Ten Times

I saw a patient with mirror syndrome this morning.

In maternal-fetal medicine, mirror syndrome is one of those diagnoses that never feels routine. Maternal edema begins to mirror fetal and placental hydrops. The mother swells as the fetus swells. The placenta is often thickened. The clinical picture can look close enough to preeclampsia to make the distinction matter.

It is rare enough to feel like an event.

But every time I see it, I have the same realization.

I have seen this before.

Not once. Several times. Across years. Across hospitals. Across EMRs that do not talk to each other and were never designed to answer the question I actually want to ask.

The condition is rare in a month.

It is not rare across a career.

That is the career registry problem.

You Are Building a Dataset and Storing It Nowhere

Over a long enough clinical life, a physician accumulates a private registry of rare things.

Mirror syndrome. Vasa previa. Hydrops from parvovirus. A fetal arrhythmia that converted. A twin transfusion case that behaved exactly like the paper said it would. Another that did not.

This is not trivia. It is an earned dataset.

It is valuable because it is local. It carries the details of the patients you managed, the systems you worked inside, the tests that arrived late, the consults that changed the plan, the ultrasound findings that did not fit neatly into the textbook paragraph.

The problem is where that dataset lives.

It lives in memory, which decays.

It lives in charts, which are fragmented.

It lives inside separate EMRs at separate hospitals, each one acting as if the career of the physician does not exist outside its own walls.

I can usually remember that a case happened. What I cannot reliably remember is the detail that helps the next patient. The hematocrit trend. The titer. The first ultrasound appearance compared with the third. The specific intervention. The time course of maternal improvement. The decision I would repeat. The decision I would not.

That detail is the asset.

That asset is exactly what fades.

The Chart Holds the Data Hostage

The chart is necessary. It is also insufficient.

It records the patient encounter for care, billing, compliance, and medicolegal continuity. It was not built to function as a physician’s longitudinal case memory.

There is no reliable button in the EMR that says:

Show me every case of fetal hydrops with maternal edema I have personally managed.

Show me which etiologies resolved after fetal treatment.

Show me the cases where the maternal picture looked like preeclampsia but the labs pointed toward hemodilution.

Show me my last ten mirror syndrome cases across every institution where I have practiced.

The information exists.

It is unreachable at the moment it would be most useful.

That is a distribution problem. The knowledge was produced. The experience was earned. But it was never placed into a system that can return it when the next patient arrives.

The Second Brain Is Not a Productivity Toy

For the last while, I have kept a second brain in Obsidian.

Plain markdown files. Local. Searchable. Mine.

When an interesting case crosses my desk, I write a short de-identified note. I tag it. I link it to related notes. I capture the clinical pattern, the decision points, and one honest line about what I would do again.

That is the whole mechanism.

The value is not in Obsidian as a product. The value is in the discipline of capture and the ownership of the record.

This is where physician knowledge management differs from generic note taking. A physician’s second brain is not a place to store interesting quotes. It is a clinical memory layer. It turns scattered experience into queryable judgment.

I think of this as the case memory layer.

It sits between the EMR and the clinician’s unaided memory. It does not replace the chart. It does not contain identifiable patient data. It does not become a shadow medical record.

It preserves the pattern the physician learned from the encounter.

One Note Per Case

For mirror syndrome, the note does not need to be complicated.

One de-identified case note is enough.

The core fields are simple:

• Presentation and gestational age
• Maternal findings that raised the concern
• Fetal and placental findings
• Suspected cause of hydrops
• Labs that helped separate mirror syndrome from preeclampsia
• Intervention or delivery decision
• Maternal response over time
• Outcome
• One sentence on what I would repeat or change

Then I add tags.

#mirror-syndrome

#hydrops

#parvovirus

#ttts

#preeclampsia-differential

The tags are not decoration. They are retrieval architecture.

The links matter too. A parvovirus-associated hydrops case connects to other parvovirus cases. A TTTS case connects to twin surveillance notes. A mirror syndrome case connects to preeclampsia differential notes because the distinction is clinically consequential.

Over time, the map forms.

Not the textbook map.

My map.

Case N Plus One

The payoff arrives when case N plus one walks in.

I am not asking my memory to reconstruct a decade of scattered encounters. I open the vault. I search one tag. Prior cases appear in seconds.

Now I can read my own past management against the current patient.

Not as a protocol. As context.

That distinction matters. The vault is not telling me what to do. It is returning the prior cases that shaped my judgment. The clinical decision still happens in the room, with the patient in front of me, using the current data.

But the decision is no longer made against a blank page.

When today’s patient arrived, I was not only remembering mirror syndrome as a rare diagnosis from training. I was remembering the last several patients in whom maternal findings tracked with fetal and placental disease. I was remembering which details changed management. I was remembering where the diagnosis became clearer and where it remained uncertain.

That changes how fast I can orient.

It also changes what questions I know to ask.

Logs Before Intelligence

I have written before about logs before intelligence.

The principle is simple. Capture the record first. Every intelligent thing you want later depends on having the record at all.

Clinical practice obeys the same architecture.

You cannot pattern-match across cases you never wrote down. You cannot query a career that was never logged. You cannot ask an AI system to synthesize your clinical experience if that experience exists only as fading memory and locked charts.

No log, no pattern.

No pattern, no compounding judgment.

This is why the second brain matters more for rare conditions than common ones. Common things are reinforced by repetition. Rare things are separated by time. The distance between cases is exactly what makes a durable external memory necessary.

The physician who seems to have an uncanny memory for rare disease may not be remembering harder.

They may have built a better retrieval system.

Privacy Is Part of the Architecture

A clinical second brain is a responsibility before it is a convenience.

The notes must be de-identified. The vault must be secured. The practice must be consistent with institutional policy, applicable law, and professional judgment.

This is not optional housekeeping. It is part of the architecture.

The value of a case memory layer is not the patient’s identity. The value is the clinical pattern: presentation, differential, decision points, response, outcome, and reflection.

That pattern survives de-identification.

If the pattern does not survive without identifiers, the note probably does not belong in a personal vault.

Start With One Case

This does not need to begin as a project.

It begins as one note.

The next time something rare crosses your desk, write the clinical pattern down after the encounter. Remove identifiers. Add one or two tags. Link it to a related topic if the connection is obvious. Stop there.

Do it again the next week.

After a few months, the structure will reveal itself because the notes will show you what you are actually collecting.

That is better than designing a perfect taxonomy before any record exists. Physicians already know this from clinical work. You cannot interpret a trend before you have serial data points.

The note is the data point.

The vault is the longitudinal record.

The Rare Will Return

The rare condition will keep walking into the office.

The only question is whether it arrives as an isolated memory or as part of a body of experience you can actually reach.

The case you saw ten years ago should not disappear because the EMR changed and your memory blurred.

Build the case memory layer.